[
Lists Home |
Date Index |
Thread Index
]
On Fri, 2004-01-23 at 18:38, Simon St.Laurent wrote:
> bob.ducharme@lexisnexis.com (DuCharme, Bob (LNG-CHO)) writes:
<snip>
> >Well, that gets back to the motive behind my original question: where
> >does a hub architecture come from, if not from a group of people
> >representing various interests gathering together to determine their
> >common interest--a committee?
>
> I'd like that committee to appear later in the process, not early.
> XLink was a poor idea because there was a notion that we already had the
> linking expertise necessary to build such a thing, even before we saw
> how the medium for these links - XML - would turn out.
You are basing your argument on the assumption that XLink works poorly,
but it does not. XLink works well for a large class of links.
XLink does not have to work for everything. It just has to work for
enough types of links to make it advantageous to share code. It does
that.
I have worked on implementing XLink systems at Ericsson, NDC, Volvo,
Scania and other companies. The linking systems work very well. I have
also been able to reuse enough code to make it worthwhile to develop it.
The biggest problems I run into are not with XLink per se, it is with a
poor understanding of the concepts of linking. That must be dealt with
no matter what linking system you want to implement.
For example, I have worked with people that seem unable to grasp the
idea of linking to a fragment to embed it during some form of
processing. Using XInclude to do this instead of XLink will not make it
easier for people grasp the basic concept of embedding a fragment, or
working with reusable chunks.
I have seen XLink designs fail. For example, one XLink design was
scrapped because the technical manager of a project wanted to use his
own horrible link scheme where links were processed in different ways
depending on how far away from each other they were in a document. This
"Twin Link" model was completely insane, but being the brainchild of the
technical manager, it won over XLink anyway. Then of course, the Twin
Links could not be implemented... In the end, a third scheme, horribly
twisted and barely usable won out. To add insult to injury, the link
elements used the XLink namespace, even though it wasn't really XLink at
all.
In my experience, the XLink projects that have failed have all failed
due to the kind of political infighting and stupidity that would kill
any project, links or no links.
> On fundamentals, I think XLink was wrong to treat 'linking' as a generic
> field of endeavor as well, so I'm not sure any single committee, no
> matter how diverse its membership, makes sense in any case.
XLink certainly isn't perfect, but from a technical point of view it
accomplishes pretty much what it sets out to do. One could certainly
argue that it could be done better, but as far as I know, to date it
hasn't.
>
> I'd rather see gatekeepers doing this work than trusting it to kings and
> councils.
I find this appealing, but the gatekeepers are at it now, and they are
making a mess of it.
>
> >To reiterate something from my last
> >post, I realize that no committee can come up with something that
> >works for absolutely everyone. But, when you have something evolving
> >on its own, completely organically, it evolves into tag soup, and
> >commercial enterprises don't want to use tag soup. They're nervous
> >enough about the differences between RSS .9, 1.0, 2.0, and Atom.
>
> Well, to get philosophical about it, the notion that there has to be one
> format, even if that format is only for interchange (which it rarely
> is), speaks volumes about our fears and very badly about our initiative.
If initiative is defined as writing the same boring implementations for
cross-references and subdocuments over and over again, then you are
right.
In every SGML/XML project I've been in the past five years, the same
linking problems have always cropped up. Every time I have to sit
through the same discussions, and use the same arguments and
counterarguments. XLink makes it possible to write a piece of software
that solves the problems, point to it and say "we'll use that!". Then
you can move on to more interesting things.
I agree that XLink does not solve every linking problem. However, I
think that the main problem isn't that XLink is bad, but that the
concept of linking itself is poorly understood.
Given the complexity of the problem, and the low level of understanding
generally, XLink does a remarkably good job.
/Henrik
|
